A comparison of limited-area 3DVAR and ETKF-En3DVAR data assimilation using radar observations at convective scale for the prediction of Typhoon Saomai (2006)

Abstract

The ensemble transform Kalman filter based ensemble 3D variational (ETKF-En3DVAR) data assimilation (DA) system is employed to evaluate the potential value of assimilating radar radial wind (Vr) data for the analysis and forecasting of Typhoon Saomai (2006). The DA system conducted cycling assimilation every 30 min when Saomai started to enter radar coverage. Within the DA cycles, the control analysis was updated by the ETKF-En3DVAR algorithm whereas the forecast ensemble perturbations in the hybrid scheme were updated by the ETKF algorithm. The benefits from the use of the flow-dependent ensemble covariance are explored by comparing the analysis increments, analysis and subsequent forecasts from the hybrid scheme with those from a pure 3DVAR using static background error covariance. Sensitivity to the horizontal correlation scale in the 3DVAR and the vertical covariance localization in the hybrid are also explored. The reduced horizontal correlation scale in the 3DVAR yields much more reasonable circulation analyses than the default scale. The vertical covariance localization scale specified in terms of geometric height instead of model levels allows for desirable spreading of Vr data to the surface. It seems that the assimilation with the hybrid method leads to further improved vortex intensity forecast and track forecast of the typhoon compared to those in the analyses from the global forecast system and 3DVAR. The results also indicated that the hybrid has a significant effect on the 12 h accumulated rainfall forecasts. Such improvements for analysis and forecast are probably due to the use of the flow-dependent background error covariance.